{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,2]],"date-time":"2026-10-02T14:54:08Z","timestamp":1790952848529,"version":"4.1.0"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":0,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/www.uspto.gov\/ip-policy\/copyright-policy\/copyright-basics"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>Published asset-pricing factors are measurements of a latent factor structure, and measurements have error. Using a spectral estimator that reads a factor's measurement-error variance off the top eigenstructure of a return panel, we grade thirteen published factors (the Fama-French five, momentum, AQR's quality, betting-against-beta and monthly-updated value, and the four q-factors) against four static test-asset universes and against conditional universes estimated by instrumented PCA and a conditional autoencoder. Each characteristic factor measures its own sort well and everything else poorly: momentum is a 97%-pure, nearly noise-free measurement on its own 25 size\/momentum portfolios and barely visible elsewhere, only the market is clean everywhere, and a placebo rules out a sorting tautology. On individual stocks, conditioning raises momentum's spanning R2 from 0.33 to 0.70-0.77, a gain that holds out of sample in both time directions, while value moves the other way. Feeding the measured noise into errors-in-variables-corrected two-pass regressions cuts the profitability premium from 5.3 to 1.7 percent per year, with a joint-bootstrap band of [1.4, 2.0]; the whole zoo can be corrected jointly only up to 16 percent of its measured noise, where the implied true-factor covariance turns singular. A point-in-time rebuild of the stock universe shows that the survivor design flattered the characteristic factors by up to 0.11 of spanning R2. A noise-injection experiment on the real panels calibrates the error bars without the simulation prior, with 90% multipliers of 2.4 to 10.5 bootstrap standard errors by universe, and no classification changes. Macro factor candidates fail the spanning screen in static and conditional spaces. Because mean-zero construction noise leaves a factor's mean unchanged and divides its t-statistic by the square root of one plus the noise variance, a factor can replicate and still be a poor measurement.<\/jats:p>","DOI":"10.2139\/ssrn.7551181","type":"posted-content","created":{"date-parts":[[2026,10,2]],"date-time":"2026-10-02T14:08:59Z","timestamp":1790950139000},"source":"Crossref","is-referenced-by-count":0,"title":["The Measurement Quality of Asset-Pricing Factors: Spanning, Noise, and the Limits of Joint Correction across Static and Conditional Test-Asset Universes"],"prefix":"10.2139","author":[{"ORCID":"https:\/\/orcid.org\/0009-0000-2439-0566","authenticated-orcid":true,"given":"Anirban","family":"Majumdar","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,2]],"date-time":"2026-10-02T14:08:59Z","timestamp":1790950139000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7551181"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":0,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7551181","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}